Analysis of Individual-Participant Data from 19 Cohort Studies

International Journal of Obesity (2020) 44:1368–1375 https://doi.org/10.1038/s41366-019-0480-3

ARTICLE

Behaviour, Psychology and Sociology Long working hours and change in body weight: analysis of individual-participant data from 19 cohort studies

1,2 3 4,5 2 5,6,7 Marianna Virtanen ● Markus Jokela ● Tea Lallukka ● Linda Magnusson Hanson ● Jaana Pentti ● 5 8,9 10,11 12 13 13 Solja T. Nyberg ● Lars Alfredsson ● G. David Batty ● Annalisa Casini ● Els Clays ● Dirk DeBacquer ● 4,8 2,8,14 2,4 10 12 15 Jenni Ervasti ● Eleonor Fransson ● Jaana I. Halonen ● Jenny Head ● France Kittel ● Anders Knutsson ● 2 16 4 5 5 4,17 Constanze Leineweber ● Maria Nordin ● Tuula Oksanen ● Olli Pietiläinen ● Ossi Rahkonen ● Paula Salo ● 10,18 6,7 6,19,20 2 6,7 Archana Singh-Manoux ● Sari Stenholm ● Sakari B. Suominen ● Töres Theorell ● Jussi Vahtera ● 21 2 5,10 Peter Westerholm ● Hugo Westerlund ● Mika Kivimäki

Received: 31 May 2019 / Revised: 1 October 2019 / Accepted: 27 October 2019 / Published online: 25 November 2019 © The Author(s) 2019. This article is published with open access

Abstract Objective To examine the relation between long working hours and change in body mass index (BMI). Methods We performed random effects meta-analyses using individual-participant data from 19 cohort studies from Europe, n =

1234567890();,: 1234567890();,: US and Australia ( 122,078), with a mean of 4.4-year follow-up. Working hours were measured at baseline and categorised as part time (<35 h/week), standard weekly hours (35–40 h, reference), 41–48 h, 49–54 h and ≥55 h/week (long working hours). There were four outcomes at follow-up: (1) overweight/obesity (BMI ≥ 25 kg/m2) or (2) overweight (BMI 25–29.9 kg/m2) among participants without overweight/obesity at baseline; (3) obesity (BMI ≥ 30 kg/m2) among participants with overweight at baseline, and (4) weight loss among participants with obesity at baseline. Results Of the 61,143 participants without overweight/obesity at baseline, 20.2% had overweight/obesity at follow-up. Compared with standard weekly working hours, the age-, sex- and socioeconomic status-adjusted relative risk (RR) of overweight/obesity was 0.95 (95% CI 0.90–1.00) for part-time work, 1.07 (1.02–1.12) for 41–48 weekly working hours, 1.09 (1.03–1.16) for 49–54 h and 1.17 (1.08–1.27) for long working hours (P for trend <0.0001). The ﬁndings were similar after multivariable adjustment and in subgroup analyses. Long working hours were associated with an excess risk of shift from normal weight to overweight rather than from overweight to obesity. Long working hours were not associated with weight loss among participants with obesity. Conclusions This analysis of large individual-participant data suggests a small excess risk of overweight among the healthy- weight people who work long hours.

Introduction working environments, the global ‘epidemic’ of obesity currently affects all age groups, all populations, and coun- Obesity is a modifiable risk factor for an array of health tries of all income levels [3, 4]. problems, including cardiovascular disease, type 2 diabetes, Psychosocial work environment may have a role in certain cancers and dementia that collectively contribute to weight control although little research has been published a substantial proportion of disease burden and death beyond perceived stress at work. Current evidence suggests worldwide [1, 2]. Resulting also from changes in living and a modest association between work stress and overweight or obesity, with limited support from longitudinal studies [5, 6]. Another work-related aspect, long working hours, Supplementary information The online version of this article (https:// may also contribute to weight gain, owing to extended doi.org/10.1038/s41366-019-0480-3) contains supplementary periods of sitting, reduced opportunities for exercise, material, which is available to authorised users. changes in eating habits leading to positive energy balance, * Marianna Virtanen and subsequently to weight gain. marianna.virtanen@uef.fi Previous studies suggest associations of long working Extended author information available on the last page of the article hours with increased incidence of cardiovascular disease Long working hours and change in body weight: analysis of individual-participant data from 19 cohort. . . 1369 and type 2 diabetes [7–9], for which obesity is also a major Germany: German Socioeconomic Panel Survey [35]; and risk factor [10, 11]. There are few prospective studies Australia: Household, Income, and Labour Dynamics in examining the link between long working hours and chan- Australia [36]. In addition, we included the Survey of ges in body weight, and the findings are inconsistent with Health, Ageing and Retirement in Europe [37], which positive [12–15], and null [16] associations reported. Fur- included data from multiple European countries (France, ther, one study found that long working hours may confer Germany, Finland, Sweden, Denmark, Slovenia, Cyprus, protection against gaining weight [17]. In addition to the Romania, Bulgaria, Greece, Czech Republic, Slovakia, paucity of evidence, studies in this field are typically Hungary, Estonia, Latvia, Lithuania, Poland, The Nether- insufficiently powered to examine weight gain amongst lands, Belgium, Luxembourg, Spain, Portugal, Italy, people who are previously not overweight. Switzerland, and Austria). Accordingly, with a large collaborative study of 19 In total, there were 122,078 participants for prospective cohorts from Europe, the US and Australia, we examined analyses who were at work and provided data on working the association between working hours and subsequent hours, sociodemographic characteristics and repeat mea- change in body mass index (BMI), including an analysis of surements of BMI. long working hours in relation to each stage of weight gain: from non-overweight to overweight or obesity, from over- Assessment of working hours weight to obesity. Long working hours may also be associated with lower probability of weight loss among those In all cohorts, information on working hours at baseline with obesity. For the first time to our knowledge, the size of was obtained from self-report. We used the previously this dataset allows to examine also this issue. used categorisation of weekly working hours: <35 (part time), 35–40 (reference group), 41–48, 49–54 and ≥55 h (long working hours) [38, 39]. In all these studies, working Materials and methods hours data were based on a continuous response, the only exception being HHS [31] where responses were based on Participants a pre-defined categorises (<30, 30–40, 41–50 and >50 h/ week). For the purposes of the present analyses, we treated We used data from 19 cohort studies (Supplementary in that study the category of 30–40 h/week as the reference Table 1) which were either taken from Individual- with <30 h/week corresponding to part time, 41–50 h/week Participant-Data Meta-analysis in Working Populations corresponding to the 41–48 h category in meta-analysis (IPD-Work) Consortium [18], or were available from one of and>50workinghourscorrespondingtothe≥55 h two digital repositories of research data (the Inter- category. University Consortium for Political and Social Research; http://www.icpsr.umich.edu/icpsrweb/ICPSR/) or the UK Assessment of body mass index (BMI) Data Service (http://ukdataservice.ac.uk/). The studies were chosen according to the availability of working hours and At baseline and follow-up, BMI (expressed as kg/m2) was BMI data in a prospective cohort design with the follow-up based on self-reported height and weight in all cohort stu- not exceeding 10 years. All studies have their own approval dies except Whitehall II and Belstress, in which height and from their relevant local or national ethics committee, and weight were measured directly during a standardised clin- all participants had given informed consent. ical examination. The following categories at baseline and Six cohort studies were from the UK: British Birth at follow-up were calculated: not overweight (BMI < 25.0); Cohort 1970 (BCS1970) [19], British Household Panel overweight (25.0–29.9); obese (30.0 or more). For those Survey [20], National Child Development Study [21], with obesity, we also determined whether they lost weight Whitehall II study [22], UK Household Longitudinal Study to non-obesity or had weight loss of >5% [40] between the [23], and English Longitudinal Study of Aging [24]; four baseline and follow-up. were from the US: Americans’ Changing Lives [25], Midlife in the United States [26], Health and Retirement Covariates Study [27], and National Longitudinal Survey of Youth [28]; three from Finland: Finnish Public Sector Study [29], All covariates were measured at the baseline and were Health and Social Support Study [30], and Helsinki Health typically either self-reported or register-based depending on Study (HHS) [31]; two from Sweden: Swedish Long- the study in question. Socioeconomic status was based on itudinal Occupational Survey of Health [32] and Work, occupational position, or educational qualification and was Lipids and Fibrinogen Norrland [33]; and one each from categorised as high, intermediate, and low. Health-related Belgium: Belgian Job Stress Project (Belstress) [34]; factors included pre-existing chronic somatic disease (e.g., 1370 M. Virtanen et al. cardiovascular disease, diabetes, asthma, rheumatoid Results arthritis), and psychological distress or depressive symptoms. Smoking (current vs non-smoker) [41] and leisure- Of the 122,078 participants, 61,143 (50.1%) were without time physical activity (low vs intermediate/high) [42] were overweight or obesity at baseline; 42,965 (35.2%) had both self-reported. overweight and 17,970 (14.7%) had obesity. Of those without overweight/obesity, 31,703 (51.9%) worked stan- Statistical analyses dard 35–40 h a week, 10,568 (17.3%) worked 41–48 h, 3897 (6.4%) worked 49–54 h and 3947 (6.5%) worked We used a two-stage approach to the present meta-analysis 55 h or more. The number of part-time workers was 11,028 [43]. First, we estimated study-specific associations with (18.0%). Of the cohort, 12,349 (20.2%) gained weight and individual participant data. With the outcomes of interest were overweight or obese at follow-up. The duration of being relatively common, we used log-binomial regression follow-up across studies was 1–9 years (mean 4.4 years), analysis with relative risk (RR) as an estimator of effect and corresponding to 267,491 person-years at risk for the its 95% confidence intervals as an indicator of precision. participants without overweight/obesity (Supplementary In the second stage, we used random effects meta- Table 1). analysis to obtain a pooled estimate from the first-stage Figure 1a–d shows the forest plots of RRs for new-onset study-specific estimates. We conducted our analyses in overweight/obesity at follow-up by the working hour cate- seven parts. Part 1: forest plots including cohort-specific gories. Compared with working standard 35–40 h a week, estimates from main analysis comparing different working working 41–48 h was associated with RR = 1.07 (95% CI hours categories in relation to weight gain—that is, transi- 1.02–1.12; Panel a); working 49–54 h with RR = 1.09 (95% tion from non-overweight to overweight/obesity), adjusted CI 1.03–1.16; Panel b) and working 55 h or more with for age, sex and socioeconomic status. Significance of trend RR = 1.17 (95% CI 1.08–1.27; Panel c). Part-time work, in was assessed by meta-regression. turn, was associated with RR = 0.95 (95% CI 0.90–1.00; Further parts were conducted comparing ≥55 weekly Panel d). These findings suggest a dose-response relation working hours to 35–40 h. Part 2: weight gain outcome as in between longer working hours and the risk of increasing analysis 1 plus adjustment for chronic somatic disease, body weight (P for trend <0.0001 in meta-regression; a psychological distress/depressive symptoms, smoking and graph presented in Fig. 2). Heterogeneity, as estimated by physical activity. Part 3: sensitivity analysis excluding I2, suggested no (a, b, and d) or moderate (c) heterogeneity smokers (adjusted for age, sex and socioeconomic status). between study-specific estimates. τ-values in Panels a–d Part 4: sensitivity analysis excluding those with chronic were 0, 0, 0.10 and 0.03, respectively. disease at baseline (adjusted for age, sex and socioeconomic Multivariable adjustment for chronic somatic disease, status). Part 5: sensitivity analysis excluding those with mental health and lifestyle attenuated the association underweight (BMI < 18.5, n = 1343; adjusted for age, sex between long working hours and onset of overweight/obe- and socioeconomic status). Part 6: subgroup analyses by sity, from 1.17 (95% CI 1.08–1.27) to 1.12 (95% CI sex, age groups (<50 years and ≥50 years), socioeconomic 1.01–1.25) (Fig. 3). Sensitivity analyses in which we status group, and by follow-up time (1-2, 3-4 and 5–9 excluded cigarette smokers, those with chronic somatic years). diseases, and those with underweight, had little impact on Part 7: further transitions between BMI categories from the associations. Similarly, results were unchanged fol- baseline to follow-up: from non-overweight to overweight lowing subgroup analyses in which we stratified the data by only; from overweight to obesity; and weight loss from sex, age, socioeconomic group, or duration of follow-up. obesity to non-obesity; and from obesity to >5% The greatest heterogeneity in effect estimates was evident in weight loss. men, among those with high or low socioeconomic status, Heterogeneity of the study-specific estimates was asses- and in studies with medium-length follow-up (3–4 years). sed using the I2 statistic and τ [44] with a higher numerical The overall PAF for overtime (>40 h) work was 2.8%, value denoting greater heterogeneity. We calculated popu- suggesting that if all full-time workers worked standard lation attributable fraction (PAF) [18, 45] for overtime 35–40 h, the population level reduction of overweight/ (>40) weekly hours, which indicates how much (in per- obesity would be 2.8%. centage) overweight/obesity in the general population Figure 4 shows the results of further analysis of the would be reduced if all full-time workers worked standard development of BMI between baseline and follow-up. Long 35–40 h, assuming that the associations are causal. The (≥55) weekly working hours were associated with weight study-specific results were computed using SAS 9.4 (Cary, gain from non-overweight to overweight (RR = 1.16, 95% NC) and the meta-analyses were conducted using Stata 15 CI 1.07–1.26 and no significant heterogeneity), but not from (College Station, TX). overweight to obesity (RR = 1.01, 95% CI 0.89–1.14). No Long working hours and change in body weight: analysis of individual-participant data from 19 cohort. . . 1371

Fig. 1 Relative risk (RR) and 95% conﬁdence interval for the asso- baseline (random effects meta-analysis adjusted for age, sex and ciation between working hour category at baseline and onset of socioeconomic status) (a–d) overweight/obesity at follow-up in participants with BMI <25 kg/m2 at

associations were observed between working hours and weight loss among participants with obesity.

Discussion

In this individual-participant data meta-analysis of 19 cohort studies with 122,078 participants, we examined the prospective associations between working hours and weight change in a greater detail than has previously been possible. We found a modest association but a suggestive dose- response relation between longer working hours at baseline and the risk of weight gain from non-overweight to overweight or obesity at follow-up. The association with weekly Fig. 2 Illustration of a dose-response pattern (trend) on the association working hours of 55 or more was further examined and the between working hours and onset of overweight/obesity from a meta- analyses suggest that the ﬁnding is robust to adjustments analysis of 19 studies and replicable across subgroups. However, our detailed 1372 M. Virtanen et al.

Fig. 3 Summary relative risk (RR) of the onset of overweight/ obesity for long (≥55 h/week) working hours at baseline compared with 35–40 h from serial adjustments and subgroup analyses in participants with BMI <25 kg/m2 at baseline. Relative risk ratios are adjusted for age, sex and socioeconomic status as appropriate

Fig. 4 Summary relative risk of weight changes from baseline to follow-up for long (≥55 h/week) working hours at baseline compared with 35–40 working hours per week

analysis of the development of BMI showed that long diseases and psychological distress, exclusion of smokers working hours may increase the risk of a shift from normal and those with chronic somatic disease or underweight, and to overweight but not from overweight to the obese cate- in subgroup analyses by sex, age, socioeconomic status and gory. Working hours were not associated with weight loss length of follow-up. Thus, although weak, the association among those who had obesity. between long working hours and weight gain appeared to be Our results are in line with previous prospective studies robust. Our sensitivity analyses suggested that we can be suggesting a positive association between long working most confident with the finding regarding a shift from hours and BMI [12–15] but in contrast with one small-scale normal to overweight category, with a RR of 1.16 with no Japanese study which reported that long working hours significant heterogeneity in study-specific estimates. were associated with lower likelihood of weight gain [17]. Possible mechanism explaining the association between The previous studies with positive findings were limited to long working hours and weight gain may be extended datasets that included older [12] or younger [13] men and periods of sitting, because today, many occupations are women, middle-aged women only [14], or industrial sedentary. Other mechanisms may be linked to too little employees [15]. With a large dataset we were able to obtain exercise and unhealthy diet, because people with long more precise estimates for the association between working working hours may not have resources, time or energy to hours and BMI, including separate analyses for men and engage in a healthy lifestyle [7]. Part-time workers— women and across age groups and socioeconomic strata. potentially having more time—seemed to have a lower risk The longest (≥55) weekly working hour category was of weight gain in our meta-analysis. This finding is in line associated with a 1.17-fold risk of new-onset overweight or with previous observations on Australian women who obesity in the model adjusted for age, sex and socio- worked part time and had a smaller risk of weight gain than economic status. The estimates from other working hour women who worked full time [14]. In our study, the 1.17- categories suggested a dose-response relationship although fold risk with long (≥55) weekly working hours attenuated the overall association was modest. The association to 1.12 in a multivariable adjusted model adding physical remained after adjustment for other lifestyle factors, chronic and mental health, smoking and physical inactivity. Long working hours and change in body weight: analysis of individual-participant data from 19 cohort. . . 1373

However, information on eating habits, one of the potential [49]. We were thus unable to assess whether the relatively underlying mechanisms, was not available in our study. small overall risk was an underestimate due to mis- Another unmeasured mechanism could be prolonged sitting classification of the exposure or outcome. However, expo- time at work, because our measures of physical activity sure misclassification may contribute to both under- and covered only leisure-time physical inactivity. However, over-estimation of associations. even though prolonged overall sitting time has been asso- The PAF was 2.8% which suggests that overweight/ ciated with a higher occupational status [46], our findings obesity could be reduced by 2.8% if all full-time workers did not suggest differences the association between long worked standard 35 to 40 h a week and the observed working hours and weight gain between socioeconomic associations were causal. Given the multifactorial aetiology groups. Moreover, commuting hours would be an important of obesity, this PAF is not trivial for population-wide pre- factor to be considered in further studies to increase our vention although the clinical significance at the individual understanding of the relationship between time-related level is modest. This size of PAF is comparable with those factors and weight gain. However, previous studies have obtained for the relation of work stress (3.4%) [18] and suggested that unhealthy lifestyle is not a major mechanism depressive symptoms (1.1%) [45] with coronary heart dis- between long working hours and cardiovascular diseases ease although direct comparisons with BMI are not avail- [7]. The modest association found between long working able. However, due to the causality assumption, findings on hours and weight gain in the present study also supports the PAF in observational studies should be interpreted hypothesis of mechanisms being other than obesity. cautiously. Long working hours were not associated with weight In conclusion, this individual participant meta-analysis of loss among participants with obesity. However, we could 19 cohort studies found a small excess risk of overweight not distinguish between intentional and unintentional among individuals who work longer hours. Detailed ana- weight loss. One could assume that psychosocial factors, lyses were consistent with the hypothesis that the impact of such as work-related stress, affect success in intentional long working hours involves the progression from normal to attempts to lose weight among overweight individuals overweight rather than from overweight to obesity and that whereas other factors, such as the onset of physical diseases working long hours is not associated with weight loss may contribute to unintentional weight loss. However, the among individuals with obesity. Future longitudinal studies relationship between stress and weight change has been assessing causality and underlying mechanisms, such as diet suggested to be complex as for some people stress seem to and accelerometer-measured physical activity would induce weight gain whereas for others it is associated with increase our understanding of the effects of long working weight loss and for still others stress has no significant hours on changes in body weight. effect on weight [6]. Important strengths of our study include the first large- Funding Era-Age2 grant, Academy of Finland (264944, 287488 and scale longitudinal individual participant data analysis of the 319200), UK Medical Research Council (R024227/1), NordForsk (the Research Programme for Health and Welfare), Academy of Finland association between working hours and weight change. A (311492), Helsinki Institute of Life Science. large sample size with harmonised variables reduces the likelihood of random error. We included cohort data from Compliance with ethical standards several countries in Europe, US and Australia, which increases generalisability of our findings, however, to those Conflict of interest The authors declare that they have no conflict of countries only. Notable limitations include an observational interest. study design, which does not allow us to determine caus- Publisher’s note Springer Nature remains neutral with regard to ality, and self-reported exposure to working hours and BMI jurisdictional claims in published maps and institutional affiliations. in almost all cohorts. Working hours was measured only once although for some employees, long reported working Open Access This article is licensed under a Creative Commons hours may have had been a temporary ‘peak’ which was Attribution 4.0 International License, which permits use, sharing, resolved shortly after the baseline survey whereas for oth- adaptation, distribution and reproduction in any medium or format, as long as you give appropriate credit to the original author(s) and the ers, excessive hours may have had been persistent. Self- source, provide a link to the Creative Commons license, and indicate if report may also involve recall bias if the participants’ recall changes were made. The images or other third party material in this hours worked inaccurately or if they overestimate their article are included in the article’s Creative Commons license, unless height and underestimate their weight [47]. However, the indicated otherwise in a credit line to the material. If material is not included in the article’s Creative Commons license and your intended validity of self-reported working hours has been found to be use is not permitted by statutory regulation or exceeds the permitted at least moderate [48] and the associations between BMI use, you will need to obtain permission directly from the copyright and cardiometabolic disease have been shown to be similar holder. To view a copy of this license, visit http://creativecommons. for self-reported and directly measured weight and height org/licenses/by/4.0/. 1374 M. Virtanen et al.

References 20. Taylor MF, Brice J, Buck N, Prentice-Lane E. British household panel survey—user manual-volume A: Introduction, technical report 1. GBD 2016 Causes of Death Collaborators. Global, regional, and and appendices. Colchester, England: University of Essex; 2010. fi national age-sex specific mortality for 264 causes of death, 1980- 21. Power C, Elliott J. Cohort pro le: 1958 British birth cohort (National – 2016: a systematic analysis for the Global Burden of Disease Child Development Study). Int J Epidemiol. 2006;35:34 41. fi Study 2016. Lancet. 2017;390:1151–210. 22. Marmot M, Brunner E. Cohort Pro le: the Whitehall II study. Int J – 2. Bray GA, Kim KK, Wilding JPH, World Obesity Forum. Obesity: a Epidemiol. 2005;34:251 6. chronic relapsing progressive disease process. A position statement 23. Burton J, Laurie H, Lynn P. Understanding society desgin over- fi fi of the World Obesity Federation. Obes Rev. 2017;18:715–23. view. Understanding society: early ndings from the rst wave of ’ 3. Guariguata L, Whiting DR, Hambleton I, Beagley J, Linnenkamp the UK s household longitudinal study. Essex: Institute for Social – U, Shaw JE. Global estimates of diabetes prevalence for 2013 and and Economic Research, University of Essex; 2011. p. 129 40. fi projections for 2035. Diabetes Res Clin Pract. 2014;103:137–49. 24. Steptoe A, Breeze E, Banks J, Nazroo J. Cohort pro le: the English – 4. Malik VS, Willett WC, Hu FB. Global obesity: trends, risk factors longitudinal study of ageing. Int J Epidemiol. 2013;42:1640 8. and policy implications. Nat Rev Endocrinol. 2013;9:13–27. 25. House JS, Lantz PM, Herd P. Continuity and change in the social fi 5. Solovieva S, Lallukka T, Virtanen M, Viikari-Juntura E. Psy- strati cation of aging and health over the life course: evidence chosocial factors at work, long work hours, and obesity: a sys- from a nationally representative longitudinal study from 1986 to ’ tematic review. Scand J Work Environ Health. 2013;39:241–58. 2001/2002 (Americans Changing Lives Study). J Gerontol B – 6. Kivimaki M, Singh-Manoux A, Nyberg S, Jokela M, Virtanen M. Psychol Sci Soc Sci. 2005;60 Spec No 2:15 26. Job strain and risk of obesity: systematic review and meta-analysis 26. Brim OG, Baltes PB, Bumpass LL, Cleary PD, Featherman DL, . of cohort studies. Int J Obes. 2015;39:1597–1600. Hazzard WR, et al Midlife in the United States (MIDUS 1), – 7. Virtanen M, Kivimaki M. Long working hours and risk of car- 1995 1996. Ann Arbor, MI: Inter-university Consortium for diovascular disease. Curr Cardiol Rep. 2018;20:123. Political and Social Research; 2018. https://www.icpsr.umich.edu/ 8. Kivimaki M, Virtanen M, Kawachi I, Nyberg S, Alfredsson L, icpsrweb/NACDA/studies/2760. Batty GD, et al. Long working hours, socioeconomic status, and 27. Heeringa SG, Connor JH. Technical description of the health and the risk of incident type 2 diabetes: a meta-analysis of published retirement survey sample design. University of Michigan: Ann and unpublished data from 222 120 individuals. Lancet Diabetes Arbor, MI. 1995 http://hrsonline.isr.umich.edu. Endocrinol. 2015;3:27–34. 28. Bureau of Labor Statistics. NLSY79. In: National Longitudinal 9. Kivimaki M, Nyberg ST, Batty GD, Kawachi I, Jokela M, Surveys, NLS Handbook 2005. United States Department of Alfredsson L, et al. Long working hours as a risk factor for atrial Labor: 2005. https://www.bls.gov/nls/handbook/nlshndbk.htm. fibrillation: a multi-cohort study. Eur Heart J. 2017;38:2621–8. 29. Kivimaki M, Lawlor DA, Davey Smith G, Kouvonen A, Virtanen 10. American Diabetes Association. Standards of medical care in M, et al. Socioeconomic position, co-occurrence of behavior- diabetes. Diabetes Care. 2015;38:S1–S94. related risk factors, and coronary heart disease: the Finnish Public – 11. Piepoli MF, Hoes AW, Agewall S, Albus C, Brotons C, Catapano Sector study. Am J Public Health. 2007;97:874 9. AL, et al. 2016 European guidelines on cardiovascular disease pre- 30. Korkeila K, Suominen S, Ahvenainen J, Ojanlatva A, Rautava P, vention in clinical practice: the Sixth Joint Task Force of the Eur- et al. Non-response and related factors in a nation-wide health – opean Society of Cardiology and other societies on cardiovascular survey. Eur J Epidemiol. 2001;17:991 9. disease prevention in clinical practice (constituted by representatives 31. Lahelma E, Aittomaki A, Laaksonen M, Lallukka T, Martikainen fi of 10 societies and by invited experts), developed with the special P, Piha K, et al. Cohort pro le: the Helsinki Health Study. Int J – contribution of the European Association for Cardiovascular Pre- Epidemiol. 2013;42:722 30. vention & Rehabilitation (EACPR). Eur Heart J. 2016;37:2315–81. 32. Magnusson Hanson LL, Leineweber C, Persson V, Hyde M, fi 12. Mercan MA. A research note on the relationship between long Theorell T, Westerlund H. Cohort Pro le: the Swedish Long- working hours and weight gain for older workers in the United itudinal Occupational Survey of Health (SLOSH). Int J Epide- – States. Res Aging. 2014;36:557–67. miol. 2018;47:691 2i. 13. Courtemanche C. Longer hours, longer waistlines? The relation- 33. Alfredsson L, Hammar N, Fransson E, de Faire U, Hallqvist J, ship between work hours and obesity. Forum Health. Econ Policy. Knutsson A, et al. Job strain and major risk factors for coronary 2009;12:1–33. heart disease among employed males and females in a Swedish fi 14. Au N, Hauck K, Hollingsworth B. Employment, work hours and study on work, lipids and brinogen. Scand J Work Environ – weight gain among middle-aged women. Int J Obes. Health. 2002;28:238 48. 2013;37:718–24. 34. De Bacquer D, Pelfrene E, Clays E, Mak R, Moreau M, de Smet 15. Lallukka T, Sarlio-Lahteenkorva S, Kaila-Kangas L, Pitkaniemi J, P, et al. Perceived job stress and incidence of coronary events: 3- Luukkonen R, Leino-Arjas P. Working conditions and weight year follow-up of the Belgian Job Stress Project cohort. Am J – gain: a 28-year follow-up study of industrial employees. Eur J Epidemiol. 2005;161:434 41. Epidemiol. 2008;23:303–10. 35. Schupp J. The Socioeconomic Panel (SOEP). Bundesge- 16. Roos E, Lallukka T, Rahkonen O, Lahelma E, Laaksonen M. sundheitsblatt Gesundheitsforschung Gesundheitsschutz. 2012; – Working conditions and major weight gain-a prospective cohort 55:767 74. fi study. Arch Environ Occup Health. 2013;68:166–72. 36. Summer eld M, Freidin S, Hahn M, Ittak P, Li N, Macalalad N, . — 17. Wada K, Katoh N, Aratake Y, Furukawa Y, Hayashi T, Satoh E, et al HILDA user manual release 12. Melbourne, Australia: et al. Effects of overtime work on blood pressure and body mass Melbourne Institute of Applied Economic and Social Research, index in Japanese male workers. Occup Med. 2006;56:578–80. University of Melbourne; 2013. https://melbourneinstitute. 18. Kivimaki M, Nyberg ST, Batty GD, Fransson E, Heikkilä K, unimelb.edu.au/assets/documents/hilda-user-manual/HILDA_ Alfredsson L, et al. Job strain as a risk factor for coronary heart User_Manual_Release_12.2.pdf. disease: a collaborative meta-analysis of individual participant 37. Börsch-Supan A, Brugiavini A, Jürges H, Kapteyn A, Mack- data. Lancet. 2012;380:1491–7. enbach J, SIegrist J, et al. First results from the Survey of Health, 19. Elliott J, Shepherd P. Cohort profile: 1970 British Birth Cohort Ageing and Retirement in Europe (2004-2007). Starting the (BCS70). Int J Epidemiol. 2006;35:836–43. longitudinal dimension. Mannheim: Mannheim Research Institute for the Economics of Aging (MEA); 2008. Long working hours and change in body weight: analysis of individual-participant data from 19 cohort. . . 1375

38. Kivimaki M, Jokela M, Nyberg ST, Singh-Manoux A, Fransson 44. Rucker G, Schwarzer G, Carpenter JR, Schumacher M. Undue EI, Alfredsson L, et al. Long working hours and risk of coronary reliance on I(2) in assessing heterogeneity may mislead. BMC heart disease and stroke: a systematic review and meta-analysis of Med Res Methodol. 2008;8:79. published and unpublished data for 603 838 individuals. Lancet. 45. Yusuf S, Joseph P, Rangarajan S, Islam S, Mente A, Hystad P et al. 2015;386:1739–46. Modiﬁable risk factors, cardiovascular disease, and mortality in 39. Virtanen M, Jokela M, Nyberg ST, Madsen IE, Lallukka T, Ahola 155,722 individuals from 21 high-income, middle-income, and low- K, et al. Long working hours and alcohol use: systematic review income countries (PURE): a prospective cohort study. Lancet. 2019. and meta-analysis of published studies and unpublished individual https://doi.org/10.1016/S0140-6736(19)32008-2.[Epubaheadof participant data. BMJ. 2015;350:g7772. print]. 40. WHO. Obesity: preventing and managing the global epidemic. 46. Bennie JA, Chau JY, van der Ploeg HP, Stamatakis E, Do A, Report of a WHO consultation. World Health Organisation Bauman A. The prevalence and correlates of sitting in European Technical Report 894. Geneva: WHO; 2000. adults - a comparison of 32 Eurobarometer-participating countries. 41. Heikkila K, Nyberg ST, Fransson EI, Alfredsson L, DeBacquer D, Int J Behav Nutr Phys Act. 2013;10:107. Bjorner JB, et al. Job strain and tobacco smoking: an individual- 47. Connor Gorber S, Tremblay M, Moher D, Gorber B. A comparison participant data meta-analysis of 166,130 adults in 15 European of direct vs. self-report measures for assessing height, weight and studies. PLoS One. 2012;7:e35463. body mass index: a systematic review. Obes Rev. 2007;8:307–26. 42. Kivimaki M, Singh-Manoux A, Pentti J, Sabia S, Nyberg ST, 48. Imai T, Kuwahara K, Miyamoto T, Okazaki H, Nishihara A, Kabe I, Alfredsson L, et al. Physical inactivity, cardiometabolic disease, et al. Validity and reproducibility of self-reported working hours and risk of dementia: an individual-participant meta-analysis. among Japanese male employees. J Occup Health. 2016;58:340–6. BMJ. 2019;365:l1495. 49. Kivimaki M, Kuosma E, Ferrie JE, Luukkonen R, Nyberg ST, 43. Riley RD, Lambert PC, Abo-Zaid G. Meta-analysis of individual Alfredsson L, et al. Overweight, obesity, and risk of cardiometa- participant data: rationale, conduct, and reporting. BMJ. bolic multimorbidity: pooled analysis of individual-level data for 2010;340:c221. 120 813 adults from 16 cohort studies from the USA and Europe. Lancet. Public Health. 2017;2:e277–e285.

Afﬁliations

1 School of Educational Sciences and Psychology, University of 11 Eastern Finland, Joensuu, Finland School of Biological & Population Health Sciences, Oregon State University, Corvallis, USA 2 Stress Research Institute, Stockholm University, 12 Stockholm, Sweden IPSY, Université catholique de Louvain (UCLouvain), Louvain- la-Neuve & School of Public Health, Université libre de Bruxelles 3 Department of Psychology and Logopedics, University of (ULB), Brussels, Belgium Helsinki, Helsinki, Finland 13 Department of Public Health, Ghent University, Ghent, Belgium 4 Finnish Institute of Occupational Health, Helsinki, Finland 14 School of Health and Welfare, Jönköping University, 5 Department of Public Health, Clinicum, University of Helsinki, Jönköping, Sweden Helsinki, Finland 15 Department of Health Sciences, Mid Sweden University, 6 Department of Public Health, University of Turku and Turku Sundsvall, Sweden University Hospital, Turku, Finland 16 Department of Psychology, Umeå University, Umeå, Sweden 7 Centre for Population Health Research, University of Turku and 17 Turku University Hospital, Turku, Finland Department of Psychology, University of Turku, Turku, Finland 18 8 Institute of Environmental Medicine, Karolinska Institutet, INSERM, U 1018 Villejuif, France Stockholm, Sweden 19 Folkhälsan Research Center, Helsinki, Finland 9 Centre for Occupational and Environmental Medicine, Stockholm 20 University of Skövde, Skövde, Sweden County Council, Stockholm, Sweden 21 Occupational and Environmental Medicine, Uppsala University, 10 Department of Epidemiology & Public Health, University College Uppsala, Sweden London, London, UK